“A study of more than 32,000 adults in 2000 found that about 23.3 percent were current smokers, the Centers for Disease Control and Prevention said. … Among those defined as current smokers -- people who had smoked at least 100 cigarettes in their lives and still smoke regularly -- 70 percent said they would like to quit.” [source] I’ve seen this statistic repeated endlessly – in press releases about the Great American Smoke-Out, for instance, and in pointed public-service ads on VH1.
But what does this statistic really tell us? Does it reflect actual preferences of real people? Does it reflect “meta-preferences,” i.e., the preferences that people would prefer to have? Why should we take these statements at face value?
In evaluating behavior, economists have traditionally relied on the notion of “revealed preference,” which means that people’s actual preferences over available options are revealed by their choices. If A and B are both available, and a person chooses A, he must prefer A to B, Q.E.D. If you continue smoking when you could quit, you must prefer smoking all things considered. More recently, behavioral economists have cast doubt on revealed preference. There may be times when people’s behavior fails to reflect their “true” preferences, or where – more provocatively – underlying internally consistent preferences don’t even exist.
I have objections to various aspects of the behavioral econ literature, but let’s say they’re right about revealed preference. There’s still a kernel of truth in revealed preference that we ought to remember: people will say all kinds of things, but what they say means very little unless accompanied by real choices, with real sacrifices and trade-offs. “Actions speak louder than words,” goes the old cliché.
Of course, speech is also a form of action. In evaluating a speech act, the revealed preference approach would conclude that the subjective benefit of speaking must be greater than the subjective cost of speaking, and no more. It would not foolishly assume the meaningfulness of what’s been said. Saying “I want X” does not reveal that I want X; it reveals that I want someone to think I want X. If the behavioral objection to revealed preference is right, then the speech act may reveal even less – but it certainly doesn’t reveal more.
If lots of people say, “I want to quit smoking,” maybe they really do wish to quit, all things considered, including the pain and difficulty of quitting. Or maybe they just know the “right” answer to the question. Quitting smoking is hard; saying you’d like to quit is easy. Ask people if they’d like to visit Jamaica, and I’ll bet most of them say yes, and they won’t be lying. But tickets to Jamaica are expensive, and talk is cheap. The real test is whether they’re buying the tickets and boarding the plane.